Algorithm Predicts Premature Births
About three decades ago, my girlfriend came to the world as a preemie, born in the 33rd week of pregnancy. She weighed only 3.7 lbs. For the record: every birth before the 37th week of gestation is considered preterm and hence risky. There were no complications, and she grew up healthily and normally. The truth is: there is a lot that could have gone wrong. There are things the doctors didn't know in the 1980ies.
One of these things is that during the last weeks of gestation, the development of the unborn's cerebral cortex takes place. Earlier this year, resesarchers of the King's College London have published a study in PNAS that shows the effect of preterm births on the developing brain: it can cause inferior cognitive performance in infants. Visual-spatial processing, decision making and working memory might all be affected. In order to assess this, the lead author, KCL's David Edwards, investigated the growth and density of nerve cells using diffusion MRI (dMRI).
Moreover, there are strong indications preterm births are a risk factor for sudden infant death, autism, ADHD and cerebral palsy. The CDC names preterm births as a risk factor for cerebral palsy, too. Breaking Bad co-star RJ Mitte suffers from a mild form of cerebral palsy. It remains unclear, though, whether his condition was caused by preterm birth or not. A recent study suggests that also exposition to phthalates (contained in food, industry and hygiene products) are one of the causes for preterm deliveries.
But hope is underway. In a recent study published in PLOS ONE, computer scientist Paul Fergus and his team of researchers have found a way to predict preterm births using machine learning and the right data. They obtained the data by harnessing electrohysterography (EHG), that is measurng electrical signals of uterine activity. This method has been used before to measure the contractions during labour.
What Fergus' algorithm does is classify the recorded signals into term and preterm signals. The trick is to measure the uterine signals early on. The paper states that "distinct contraction-related, electrical uterine activity is present early on in pregnancy, even when a woman is not in true labour." The amount of these signals increases steadily during the pregnancy and shoots up during the last three to four days.
The new method is still prone to errors and produces results of varying quality, depending on the underlying data. The number of examined records suggests Fergus was not exactly using big data. 300 records of which only 38 were preterm records. Fergus himself states in the conclusion of his paper, that the low number of preterm records didn't allow the machine learning classifiers to learn properly.
Even though, the concept works and seems promising. Once all errors have been eliminated, the technique could predict a "Yes" or "No" to answer the question "Will my child be born on term?" The resarchers suggest to further improve the quality of the algorithm's outcome in the future: The answer could then be how many weeks remain until the birth.
Pair this with the right countermeasures and it might increase the odds the newborn survives. In 2012, the Daily Mail reported about a pacemaker that prevents muscles from contracting by running electric bursts into them. If applied at the right time, this might defer the birth. However, the Daily Mail didn't name any sources, studies or persons, so it remains a mystery whether such a device exists or not.
Other precautions are proven and exist already. For example, most preemies don't have fully developed lungs, which is commonly treated with corticosteroids.
This was also the case for my girlfriend.
Knowing when the premature birth will happen could lead to better preparation and hence a preciser treatment immediately after the baby is born.
Now, let's develop the idea a little further and imagine a tricorder equipped with the necessary sensors, electrodes, sufficient computing power and the right algorithms to calculate the probability of a preterm birth within seconds. Prototypes of tricorders exist already, such as the Scanadu Scout that measures heart rate, blood pressure etc. just by holding it against your skin.
The March of Dimes' premature birth report shows that the US preemie birth rate has been falling since 2006. The report also depicts that risk factors such as the absence of health insurances or smoking are continually being reduced.
Maybe technology can contribute to reduce that rate further.
D’Onofrio BM, Class QA, Rickert ME, Larsson H, Långström N, Lichtenstein P. Preterm Birth and Mortality and Morbidity: A Population-Based Quasi-experimental Study. JAMA Psychiatry. 2013;70(11):1231-1240. doi:10.1001/jamapsychiatry.2013.2107. http://archpsyc.jamanetwork.com/article.aspx?articleid=1743009
Fergus P, Cheung P, Hussain A, Al-Jumeily D, Dobbins C, et al. (2013) Prediction of Preterm Deliveries from EHG Signals Using Machine Learning. PLoS ONE 8(10): e77154. doi:10.1371/journal.pone.0077154. http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0077154
Schecter A, Lorber M, Guo Y, Wu Q, Yun SH, Kannan K, Hommel M, Imran N, Hynan LS, Cheng D, Colacino JA, Birnbaum LS. 2013. Phthalate Concentrations and Dietary Exposure from Food Purchased in New York State. Environ Health Perspect 121:473–479; http://dx.doi.org/10.1289/ehp.1206367.
The March of Dimes Premature Birth Report Card. http://www.marchofdimes.com/glue/files/premature-birth-report-card-united-states.pdf
WHO Preterm Birth Fact Sheet: http://www.who.int/mediacentre/factsheets/fs363/en/